Outcomes of implementing a coaching-based WHO Safe Childbirth Checklist program in India

Background Facility-based childbirth in low-resource settings has increased dramatically over the last two decades, yet quality of care gaps persist and mortality rates remain high. The World Health Organization (WHO) Safe Childbirth Checklist, a quality improvement tool, promotes systematic adherence to practices known to save lives and prevent harm during childbirth. Methods We conducted a matched-pair, cluster-randomized controlled trial in 60 pairs o facilities across 24 districts of Uttar Pradesh, India to test the effectiveness of the BetterBirth program, an 8-month coaching-based implementation of the Checklist, on a composite outcome of 7-day maternal/perinatal mortality and maternal morbidity. Outcomes—assessed 8-42 days post-partum—were compared between study arms adjusting for clustering and matching. We also compared birth attendants’ mean adherence to 18 essential birth practices in 15 matched pairs of facilities at 2 and 12 months after intervention initiation. Results Of 161,107 eligible women, we enrolled 157,689 (98%) and determined 7-day outcomes for 157,145 (99.7%) mother-newborn dyads. Of 4888 observed births, birth attendants’ adherence to practices was significantly higher in the intervention (I) than control (C) arm (I: 73% vs. C: 42% at 2 months, p≤0.01; I: 62% vs. C: 44% at 12 months, p≤0.01). However, we found no difference in the composite outcome (I: 15.1% C: 15.3%, RR: 0.99, 95% CI: 0.83-1.18, p=0.90). Conclusion The coaching-based WHO Safe Childbirth Checklist program produced increased adherence to some essential birth practices, but did not reduce morbidity and mortality. (Clinical Trials #NCT02148952; The Bill & Melinda Gates Foundation)


Site selection and matching
Site selection from the 320 potential sites across 38 districts was narrowed to 120 sites through the matching process. Thirty-seven facilities did not meet the inclusion criteria for study participation; those criteria were facilities with: (1) a designation as a Primary, Community, or First Referral Unit Health Center; (2) ≥1000 deliveries annually; (3) ≥3 birth attendants trained as auxiliary nurse midwives (or higher); (4) no other concurrent quality improvement or research programs; and (5) district and facility leadership willing to participate.
Each facility, meeting the above criteria, was matched to another facility using the following parameters: (1) geographic zone/location; (2) same classification/type (note community health centers and first referral units were considered equivalent); (3) average delivery load +/-500 deliveries; (4) staffing total +/-2 attendants; (5) closest single facility on distance to district hospital. All facilities were run through the matching algorithm and matches were set aside. The algorithm was then again run on nonmatched facilities. We did not relax our matching criteria to find additional matches.
Rather, we used all available matched and selected the first 60 pairs. We compared sites that were matched and enrolled in the trial compared to the sites that were not matched. There was a significantly higher delivery load in non-matched sites compared to matched sites (p<0.01), but other matching characteristics were similar between matched and no-matched sites.

Matched Sites
Non

Sample Size Calculation
In the design phase of the trial, sample size calculations were based on the assumption that the event rate of the composite outcome would be 60 events per 1000 births. The primary outcome in this trial was the rate of the composite measure of maternal death within 7 days, severe maternal complications within 7 days, fresh or macerated stillbirth, and neonatal death within 7 days. The challenge we faced in determining a baseline rate for the composite outcome is outlined in the BetterBirth Protocol. In 2011, data available from Uttar Pradesh suggested that the baseline rate of the primary outcome would be no more than 50-60/1,000 live births. Preliminary data from the study collected during the pilot phases suggest that the baseline rate may actually be greater than 100/1,000. The baseline rate used in the sample size calculations was set at 60 events/1,000. This was purposively set lower than the pilot data to (1) Table 2 in the manuscript.
Of the practices measured, 18 are key behaviors that should be conducted or essential supplies that should be available for all women and newborns during childbirth. We created a summary score of adherence to those key 18 practices. We calculated the score as a percent adherence out of the 18 practices; we calculated the average (mean) of that adherence level and reported it as a proportion. The behaviors and supplies were not conditional on appropriate administration. For example, provision of magnesium sulfate is not included in this summary score of 18 as those medicines should only be provided in specific cases.

Data Safety and Monitoring Board
A Data Safety Monitoring Board (DSMB), consisting of Dr. Cyrus Mehta (Chair), Dr.
Shally Awasthi, and Dr. Nozer Sheriar, met every 6 months after enrollment initiation and conducted an interim-analysis review when 30% of data was collected. For the interim analysis, the Haybittle-Peto 1 stopping rule was used with a p-value <0.001 for our composite measure comparing study arms. In addition, the DSMB reviewed women's complication rates and birth attendants' adherence to the Checklist. With the Haybittle-Peto approach, it is appropriate to use the conventional significance level of p-value <0.05 in the final analysis.